In 2026, the smartphone market exhibits a clear divergence pattern of "declining volume and rising prices." According to IDC data, the global average selling price (ASP) of smartphones reached USD 581 in 2026, a year-on-year increase of 27.6%, indicating that the industry as a whole continues to concentrate on the mid-to-high-end flagship market. Against this backdrop, on September 16, nubia, in collaboration with Doubao, officially released the iterative mass-produced model NaviX Ultra. Priced in the CNY 6,000 tier and positioned as a high-end flagship, it is also the industry's first Doubao AI agent phone targeted at mass commercial use.
Looking back at the Doubao phone, in December 2025, nubia launched the first-generation Doubao phone M153 technical preview version, priced at CNY 3,499. The initial batch of 30,000 units sold out rapidly, once sparking heated industry discussions.
A year later, the NaviX Ultra was officially approved for network access and launched, completing the iterative upgrade from a technical preview device to a consumer-grade mass-produced device, while also achieving comprehensive upgrades in compliance, architecture, and interaction methods.
However, the current development of the AI agent phone market is not ideal, and it is difficult to achieve large-scale popularization in the short term due to practical issues such as the lack of rigid mass demand. MIIT Smart Media (Communications World) interviewed Chen Fengwei, Deputy General Manager of China Unicom Huasheng. He believes that at this stage, the lack of universal rigid demand and the absence of strong user dependence are the fundamental reasons why agent phones are difficult to scale. Ordinary consumers cannot intuitively perceive the actual value brought by agents, and the products fail to solve high-frequency pain points for the masses. They can only be confined to geeks, office workers, and young demographics, making it difficult to achieve market scale in a short period.
Agent Phones Remain in the Exploration Stage, Making Large-Scale Adoption Difficult
The CNY 6,000-tier agent flagship precisely corresponds to the market trend of rising industry ASPs and concentration in the high-end segment. However, agent phones as a whole still belong to niche products, and there is still a long way to go before universal popularization.
In Chen Fengwei's view, the core audience for agent phones is currently very clear, mainly divided into two categories: first, young students who are willing to accept new technologies and pursue innovative experiences; second, professional groups with professional device operation capabilities who frequently use office agents. At this stage, office agents have the most implemented scenarios and the most mature development, which is also the core direction for long-term breakthroughs in mobile agents in the future.
Currently, the biggest shortcoming of the industry lies in the lack of generalization capabilities. Chen Fengwei cited an example that the most popular agent applications across the network are basically concentrated in the office vertical field. The daily active users (DAU) of leading office agents are only 20 to 30 million. Compared to the domestic base of 1.8 billion mobile phone users, the penetration rate is extremely low. As a vertical scenario tool, office agents cannot cover the diverse daily needs of ordinary users. Therefore, agent phones have not yet formed a mature mass popularization model and remain in the overall stage of exploration and trial and error.
Industry Compliance System Perfected; Security and Privacy Are No Longer Development Bottlenecks
When the first-generation Doubao phone was released, it raised issues such as data security and user privacy. With the rapid catch-up of the domestic regulatory system and the swift implementation of standards, relevant national competent departments have introduced special certification standards, testing specifications, and access systems for on-device large models and terminal agents.
On July 15, 2026, the Cyberspace Administration of China (CAC) published the first batch of filing lists for on-device generative AI services for mobile phones. Seven manufacturers, including nubia, Huawei, Apple, and Xiaomi, were all included in the compliance filing system, marking the official entry of mobile agents into a standardized, regulable, and compliant era.
Chen Fengwei stated that after more than a year of policy refinement and industry rectification, a complete security framework has been formed, whether for large model algorithms, on-device agent scheduling, mobile terminal data processing, or permission invocation. Currently, industry security, privacy, and compliance risks have been basically cleared to zero. Consumers and the industry no longer need to worry about security issues, and security is no longer an obstacle restricting industry development.
Exploring Users' Rigid Demand Is the Core of AI Phones Breaking into the Mainstream
Compared to security issues, Chen Fengwei emphasized that the lack of universal rigid demand and the absence of strong user dependence are the fundamental reasons why agent phones are difficult to scale. Ordinary consumers cannot intuitively perceive the actual value brought by agents, and the products fail to solve high-frequency pain points for the masses. They can only be confined to geeks, office workers, and young demographics, making it difficult to break into the mainstream market.
The development of AI phones is divided into three stages.
The first stage is application-level and tool-level AI phones, which involve embedding AI functional apps in traditional phones, where AI serves only as an additional tool. Currently, more than half of the phones in China belong to this category, which are also AI phones in a broad sense.
The second stage is system-level AI agent phones. Through deep transformation of the operating system and hardware underlying layers, it achieves deep integration of AI with the whole device. This is also the current mainstream iteration direction in the industry, and the newly released NaviX Ultra belongs to this stage of products.
The third stage is AI-native phones, which is the ultimate form of the industry. The hardware of the whole device and the underlying system are entirely designed for AI natively. AI becomes a basic attribute of the phone, thoroughly completing the replacement of traditional phones.
Regarding the new Doubao phone, Chen Fengwei evaluated that the new device adds API and MCP standardized interoperability protocols based on the GUI of the first-generation product, solving the risk control conflicts and compliance issues of the first generation. However, it is overall an improved iteration without revolutionary innovation. As the number of competing products in the market increases, its market appeal will gradually decline.
Closed Ecosystems; Industry Development Thinking Urges Upgrading
At the industrial ecosystem level, Chen Fengwei frankly pointed out that leading AI manufacturers and mobile phone brands currently lack a win-win ecosystem thinking. Product ecosystems are highly closed, and cross-platform interoperability is extremely poor. The industry still remains in the old logic of "monopoly of a single entry point" rather than joining hands to build a new ecosystem.
Taking operators as an example, they are actively embracing the new track of AI agents and proactively seeking collaborative construction. However, terminal manufacturers have a weak willingness to open up, and upstream and downstream linkage is insufficient, making it difficult for the new industrial ecosystem to take shape for a long time.
Vertical Scenarios First: The Most Feasible Commercialization Path at Present
Under the current situation where universal scenarios for the general public are difficult to implement, Chen Fengwei proposed that agent phones should prioritize deep cultivation in vertical professional scenarios, achieving small-scale closed loops with tool attributes.
Occupational groups with high-frequency hand-restricted work, such as food delivery riders, ride-hailing drivers, and domestic workers, are the user circles most compatible with agent capabilities. Through exclusive functions such as full-voice intelligent interaction, automatic message replies, delivery reminders, and itinerary broadcasting, agent phones can be upgraded from entertainment consumption devices to professional production tools. These users generally own multiple phones and have clear tool usage needs, making it easier to form usage stickiness and willingness to pay, and able to replicate the commercial closed loop of office agents.
He cited office agents as an example: the core of office agents being able to run through the business model lies in sufficient data, user matching, and clear value. The PC end has accumulated massive structured office data, office workers possess mature operation capabilities, AI tools can replace a large amount of repetitive work, and users are willing to pay for efficiency improvement.
In contrast, on the mobile end, effective structured data is scarce, privacy data cannot be commercialized, and users default to AI functions being free. C-end general-purpose mobile phone agents currently do not yet possess mature commercialization conditions.
Conclusion
The reshaping of the mobile phone industry by AI is not about single-product iteration, but ecosystem reconstruction. Future competition among mobile phone terminals will no longer be a contest of hardware parameters and single-model capabilities, but a competition of full-domain ecosystem systems.
He finally added, "Overall, AI agent phones represent a certain future for the smartphone industry, but the industry is still in the early stage of gradual penetration and slow replacement. Only when mobile phone manufacturers abandon closed thinking, jointly build a win-win ecosystem, and implement real rigid demand scenarios can agent phones achieve large-scale popularization and promote the terminal industry to complete a new round of AI transformation."
Editor-in-Charge / Layout: Zhu Wenfeng
Reviewed by: Mei Yaxin
Supervised by: Liu Qicheng